Copying of Unaligned Data in a Pipelined Operation

1. A method of copying unaligned data comprising: performing a pipelined loop by overlapping execution of instructions in a repeated sequence of operations, wherein an iteration of the pipelined loop includes: loading in a pipelined operation unaligned data from a string of unaligned data to form units of unaligned data, wherein loading includes: loading into a first temporary data register, a remainder of unaligned data; loading into a second temporary data register, a first unit of unaligned data from the string of unaligned data; and loading into a third temporary data register, a second unit of unaligned data from the string of unaligned data; shifting in a pipelined operation portions of the units of unaligned data to form aligned portions of data; merging in a pipelined operation the aligned portions of data into units of aligned data; storing in a pipelined operation the units of aligned data to form a string of aligned data; and moving into the first temporary data register, data from the third temporary data register, to become the remainder on a next iteration of the pipelined loop; and wherein the method includes executing one iteration of the pipelined loop in one processor cycle.

2. The method of claim 1 , wherein the shifting includes: shifting into a first shift data register, data from the first temporary data register, shifted by a shift amount; shifting into a second shift data register, data from the second temporary data register, shifted by a shift complement amount.

3. The method of claim 2 , wherein the shifting includes: shifting into a third shift data register, data from the second temporary data register, shifted by the shift amount; and shifting into a fourth shift data register, data from the third temporary data register, shifted by the shift complement amount.

4. The method of claim 1 , wherein the merging includes merging into a first merge data register, data from a first shift data register merged with data from a second shift data register.

5. The method of claim 4 , wherein the merging includes merging into a second merge data register, data from a third shift data register merged with data from a fourth shift data register.

6. The method of claim 1 , wherein the storing includes storing into a first store data register, data from a first merge data register.

7. The method of claim 6 , wherein the storing includes storing into a second store data register, data from a second merge data register.

8. A computer readable medium having instructions for causing a pipelined machine to perform a method comprising: performing a pipelined loop by overlapping execution of instructions in a repeated sequence of operations, wherein an iteration of the pipelined loop includes: loading in a pipelined operation unaligned data from an unaligned data item to form units of unaligned data, wherein loading includes: loading into a first temporary data register, a remainder of unaligned data; loading into a second temporary data register, a first unit of unaligned data from the unaligned data item; and loading into a third temporary data register, a second unit of unaligned data from the unaligned data item; positioning in a pipelined operation portions of the units of unaligned data to form aligned portions of data; merging in a pipelined operation the aligned portions of data into units of aligned data; storing in a pipelined operation the units of aligned data to form a string of aligned data; and moving into the first temporary data register, data from the third temporary data register to become the remainder on a next iteration of the pipelined loop; and wherein the method includes executing one iteration of the pipelined loop in one processor cycle.

9. The medium of claim 8 , wherein the method includes: performing the loading as a first set of pipelined tasks; performing the positioning and the merging as a second set and a third set of pipelined tasks; and performing the storing as a fourth set of pipelined tasks.

10. The medium of claim 8 , wherein the method includes: performing the loop wherein the positioning includes: positioning into a first rotating position data register, data from the first temporary data register, positioned by a position amount; positioning into a second rotating position data register, data from the second temporary data register, positioned by a position complement amount; positioning into a third rotating position data register, data from the second temporary data register, positioned by the position amount; and positioning into a fourth rotating position data register, data from the third temporary data register, positioned by the position complement amount; and rotating the rotating position data registers each time the loop is performed.

11. The medium of claim 10 , wherein the method includes: performing the loop wherein the merging includes: merging into a first rotating merge data register, data from the first rotating position data register merged with data from the second rotating position data register; merging into a second rotating merge data register, data from the third rotating position data register merged with data from the fourth rotating position data register; and rotating the rotating merge data registers each time the loop is performed.

12. The medium of claim 11 , wherein the method includes performing the loop wherein the storing includes: storing into memory pointed to by a first store data register, data from a first rotating merge data register; and storing into memory pointed to by a second store data register, data from a second rotating merge data register.

13. A computer readable medium having instructions for causing a device to perform a method, comprising: performing a pipelined loop by overlapping execution of instructions in a repeated sequence of operations, wherein an iteration of the pipelined loop includes: loading as a first set of pipelined tasks unaligned data from an unaligned data item to form units of unaligned data, wherein loading includes: loading into a first temporary data register, a remainder of unaligned data; loading into a second temporary data register, a first unit of unaligned data from the unaligned data item; and loading into a third temporary data register, a second unit of unaligned data from the unaligned data item; shifting as a second set of pipelined tasks portions of the units of unaligned data to form aligned portions of data; merging as a third set of pipelined tasks the aligned portions of data into units of aligned data; storing as a fourth set of pipelined tasks the units of aligned data to form a string of aligned data; and moving into the first temporary data register, data from the third temporary data register to become the remainder on a next iteration of the pipelined loop; and wherein the method includes executing one iteration of the pipelined loop in one processor cycle.

14. The medium of claim 13 , wherein the method includes performing the loop until all of the sets of pipelined tasks are false.

15. The medium of claim 13 , wherein the method includes: performing the loop wherein the shifting includes: shifting into a fourth temporary data register, data from the first temporary data register, shifted by a shift amount; shifting into a first rotating shift data register, data from the second temporary data register, shifted by a shift complement amount; shifting into a second rotating shift data register, data from the second temporary data register, shifted by a shift amount; and shifting into a third rotating shift data register, data from the third temporary data register; shifted by a shift complement amount; and rotating the rotating shift data registers each time the loop is performed.

16. The medium of claim 15 , wherein the method includes: performing the loop wherein the merging includes: merging into a first rotating merge data register, data from the fourth temporary data register merged with data from the first rotating shift data register; merging into a second rotating merge data register, data from the second rotating shift data register merged with data from a third rotating shift data register; and rotating the rotating merge data registers and the rotating shift data registers each time the loop is performed.

17. The medium of claim 16 , wherein the method includes: performing the loop wherein the storing includes: storing into memory pointed to by a first store data register, data from a first rotating merge data register; and storing into memory pointed to by a second store data register, data from a second rotating merge data register.

18. The medium of claim 13 , the method including performing the loop in a sequence.

19. A computing device comprising: a processor; a memory, connected to the processor; program instructions storable in the memory and executable by the processor to: perform a pipelined loop by overlapping execution of instructions in a repeated sequence of operations, wherein an iteration of the pipelined loop includes pipelined operations to: load unaligned data from a string of unaligned data to form units of unaligned data, including operations to: load a remainder of unaligned data into a first temporary data register; load a first unit of unaligned data from the unaligned data item into a second temporary data register; and load a second unit of unaligned data from the unaligned data item into a third temporary data register; shift portions of the units of unaligned data to form aligned portions of data; merge the aligned portions of data into units of aligned data; store the units of aligned data to farm a string of aligned data; move into the first temporary data register, data from the third temporary data register to become the remainder on a next iteration of the pipelined loop; and execute one iteration of the pipelined loop in one processor cycle.

20. The device of claim 19 , including pipelined program instructions storable in the memory and executable by the processor to perform the pipelined operations in a loop.

21. The device of claim 20 , including pipelined program instructions storable in the memory and executable by the processor to perform the loop until the entire string of unaligned data is loaded, shifted, merged, and stored.